> > As I now have a G400 I'd like to give it a try.

I need to check the card in the machine next door. IIRC it's also a Gx00.
Maybe I also give it a try.

> Dualhead I believe is documented in the regular G400 docs. I haven't had a
> look at this, since well, I don't have a G400 for one, and what I have heard
> hasn't been pretty from a drivers writer perspective. Discussions on other
> lists I follow have led me to believe that mutlihead is availible with a
> number of restrictions. You have to use 16 or 32 bpp modes, no others are
> availible (I don't know if you can have different layouts on different
> monitors, but I think you can) your max dclk is split in half, ie, from 310
> to 155 on most cards. 

Sure. The memory bandwidth is limited, so you need to adapt. I assume you
can run it in other split ratios as well. The limiting factor is basically
the total memory bandwidth.

> also, from a programming perspective, you still have one card, one memmory, 
> one accel engine, one warp engine, but two crtc's (including extended crtc 
> registers) 

O.K. - this is very similar to the VideoOut interface of my PM2 card.

The point is: How should we manage that ?

One the one hand, one can program it to be a completely independent head, by
placing the buffers in separate memory areas, but OTOH, it can be used as 
Zoom-In or other stuff into convetional memory, or it can display the result
from the VidIn part ...

> I believe we sould need a way for kgi to manage onboard memory even for 
> the framebuffers here, 

That depends ... Should we export that as multiple framebuffers ?

I would actually rather propose to make it a "subbuffer", as applications
might want to control it dependent on the main fb ...
We could then make something like a "subbuffer-target", that acts like any
metatarget and relays commands to such a subbuffer.

> and such, and I'll put the code up on http://194.218.183.235/~wlfshmn/
> within a couple of hours, since I have some biology homework to help my
> sister with :)

Your sister ? Hmm ... seems to be real homework, then ... :-).

CU, Andy

-- 
= Andreas Beck                    |  Email :  <[EMAIL PROTECTED]> =

Reply via email to